00:00
54, 3.5. 21. Awesome. Welcome, everybody. Thank you so much for coming to how to achieve cloud efficiency with pure fusion. Hopefully, you're here because you're wondering about better ways to manage your storage as a raise count grows like crazy. You're thinking about bringing cloud storage
00:20
concepts on premises. Not on premise, but on premises. Right. That's what we're gonna walk through in this session a little bit. Uh, my name is Andrew Miller. I won't do that again. My name is Andre Miller, lead principal technologist for America's, and I am joined by my peer.
00:34
Hi, I'm Savas Niclas. I'm the principal technologist for em. So we've been, uh, working a lot together the last four or five months. Uh, I have been embracing AD M meetings as I'm on the East Coast, and Savas is in London. So it's a It's a nice five hour time difference. It's not too bad. I've been collaborating a lot.
00:49
Now what we want to go into today is a relatively lightweight agenda where we're gonna walk through not actually starting with just really What? What? Fusion is but a little bit going further than that because some of these concepts, the things we're going to talk about are things that I wanted we wanted. Maybe all of you have wanted your careers for a
01:07
very long time about how we manage things about cloud concepts on premises. It was easy. It would already be done. Right. So we're gonna start off with a little bit of intros more for ourselves. More about why we care about this topic based on some of our background. Frankly, then moving into, we're gonna start at the end.
01:23
Actually, what are we hoping to accomplish here? What are you potentially hoping for? Or maybe if you came to the session not really know what you're looking for? Maybe a little bit of like, Why do you care about this topic to begin with, then move into some building blocks because there have been attempts at some of these concepts in the past and they haven't worked out, and it's easy to get a little bit jaded
01:42
like, Oh, these are big claims. And I don't think they could be true this time because they weren't true five years ago or 10 years ago. So what were the things that didn't work in the past and what has changed since then? They're moving into two sets of challenges. We we tried to put ourselves into your shoes. I almost kind of sitting at your your virtual desk, if you will think about the challenges
01:58
that are managing up. If you're a manager director, you're thinking about managing up within your organisation, to your boss or the CIO, or maybe managing out to the application stakeholders, people outside your organisation. And then also managing down or managing internally like down is like bad, right? It's just you're a director, you have a manager,
02:13
you have employees you have admins or even managing down the infrastructure of the pieces you have to deal with. And then last But not least, there has to be a money aspect to this. The total cost of ownership. So some quantification around the benefits of maybe we want to do all these things. But can we actually justify the mat and work in the real financial business world that we all
02:30
live in? So first intros to us. Cool. Thanks, Andrew. Yes, a little bit about me. I've had, um I lead principal technologist for me As you know, I've been in the storage industry 23 years, and and most of that time my focus has been global accounts and large enterprise customers. And obviously there's a really good fit there
02:50
with what fusion can offer. Hence, this session, Um, I've been at pure quite a while. So back nine years ago, it's when I started when we were about 400 employees. 6000 now. So and over that time helped a lot of global and large customers understand their storage needs and help them with their challenges.
03:12
Andrew and lead principal technologist at America Um, for better or worse and maybe most importantly, today start on the customer side. Actually, we mapped similar areas in our careers. We kind of admin engineer to architect. Um, I was the guy who wrote Crazy long scripts to provision hundreds of thousands of 10
03:28
gigabyte runs to a IX for DB two, whether that was needed or not, right. And then you get into different kinds of provision at a network level, or even management sometimes at scale. Eight years on the partner side, you know, wave your hands around, make promises on the white board come through, you get to come back.
03:41
If not, not so much a kind of a manager director there. I built out a technical marketing team for a company called Rubric while they were in hyper growth mode. So I've been orbiting around some of the questions of management and scale and operations for a while. And over the years I've tried not to personal comment kind of lose what I think of as my
03:57
technical soul. We all grow up and we all be sure in various ways. But for a lot of us, at least I think we got into this field because we enjoy it. We love the technology, but we need it to be relevant to the business. They are our customers and relevant externally, right, that kind of thing. So, like I mentioned,
04:13
I think I'm continuing. Let's start at the end, so desired. End state. Ah, right, yeah. So when we thought about this session, we thought, OK, let's start with the end state. What if we could have a number of things? What if we could have an environment where the users and consumers of storage got an
04:32
experience that's very similar to public cloud. Uh, you know, the cloud operating model that we're here about. And, um, and so you know, what if those users could have invisible hardware? Could have that cloud experience could have self service if we flip it around from a provider standpoint. So the idea of the consumers,
04:52
maybe the providers, internally at it. There's a lot of stuff that we have to grapple with there that we often don't want to. We would prefer for things to be policy driven. Who here has heard policy driven for, I don't know, a decade or two, and it's only kind of true, right? We actually want to encapsulate things and
05:08
policies, or whether that's around storage or protection. As environments grow and mutate, we need to be able to move things around because a raise or physical assets get hot or they get cold or things grow unpredictably. That never happens, that kind of thing, and then even the day to day management aspect, the level of operational overhead. We want to make things better and more
05:26
automated, but we don't want that automation to lead to fragility. I'm not sure if anyone else has had that you automate a bunch of things and the business requirements are too fluid. And then all the work that you put into it, it's not flexible enough, and it basically falls down on itself. Yeah, so there's external pressures. Agility is a is a key requirement from users.
05:46
The other one we'll come on to a bit later is storage is growing at an unprecedented rate. How do you cope with that? And then in the modern world, users don't really care about infrastructure. So we care about infrastructure. We're all here, but your users may not.
06:01
And then how do you How do you handle that? Demand for resources is expected now via an API or a few clicks. That's table stakes. And then, if you add if you want features, how do you add those instantly and very quickly, or deploy those quickly and then, lastly, that desire for capacity is seemingly in infinite.
06:24
How do you capacity plan or cope with that unpredictability and that growth and that desire for infinite capacity? So there's external pressures. If anything, these have accelerated thanks to the consumerization of IT. Um, there there's often internal pressures.
06:40
I, I wanna pause here for a second, if you don't mind. I'm curious. How many folks here if we if we kinda separate ourselves into two categories Uh, apologies there. But how many people would say that you are of the two options? You are more technical with business, focus and aptitude or you were more business focused with
06:57
technical awareness and knowledge. So if I don't mind to show of hands here, how many folks you put yourself more kind of in the technical camp? Curious. Maybe a little more than half. How many folks are more kind of in the business game? It doesn't have to be OK, It doesn't have to be like we We could even be blended in hybrid
07:10
based on what meeting you're in. So you're meeting with some people are like, Oh, I'm the most technical person here. And then other people are like, Oh, I'm the most business person here. You know, kind of thing. We can all we can all switch roles. OK, so some of the external pressures these are often generated externally from a business
07:22
perspective, the internal pressures, this is what we've had to live with for a while. We've had to live with pools of storage that are managed individually and separately. And as things grow, it adds to the overhead. When we provision this, If you've all heard the classic thing, asking a storage request takes days or weeks or months, right, because the level of manual intervention that has to happen and that the break points
07:42
where humans have to insert themselves to resolve things, how often or do requirements stay static for the entire life of a project? Maybe never, right? We just don't know when or how, or how much right kind of thing. So predictable, unpredictable performance requirements, much less that every time that we grow and expand,
08:00
we're actually adding future amounts of unknown, unpredictable untime technical debt, a K, a hardware refreshes or scale limitations. We know there are going to be limits. We just don't know when they're going to run into them, and so this can be a little bit of almost like you don't want. You don't want to know what you don't know, but you know it's gonna be painful.
08:17
Someday you just hope it's for the next guy. Maybe not you or the next lady. Sorry, as as like maybe I'll be done and I'll be off to the next thing and someone else lets us deal with the limits. I don't know why I'm gonna hit them, but they're coming. So what are the building blocks here? Make sure Is there anything I left out you want to highlight?
08:32
Good. We're good. We practised a little bit. But not so much reading off a script. That was the goal, you know. So building blocks, it's I'm not sure. Folk. I'm guessing I'm just gonna blatantly stereotype here. Some folks here have been doing some of this
08:46
for as long as I have 15, 20 years or so, but maybe longer, maybe less. OK, so there have been in the past where we've wanted this, and actually, I should leave that there, so it's easy to be a little bit jaded. So we wanted to talk about what didn't work in the past and what's changed today. Especially with pure to make some of this possible.
09:03
Like solving these problems in better ways. Um, personally, just gonna be very real. I've worked with storage virtualization technologies that are in band. They might have plex in the name I've worked with ones that were, uh, named after. I don't know, snakes. If I can say it that way, like I there was one
09:18
called Viper out there in the day. Um, sometimes they have to be in band. Sometimes they're out of band. What we saw, what I saw over the years, there were challenges that were more about the underlying components and the maturity. So the stuff that classically was challenging there wasn't a P completeness,
09:33
the ability to fully automate or even in some cases, a P maturity. If you're familiar with Swagger Project that's moved into open API the sense of like, even when I was a customer and I went to automate something, and could I do it all, or would I have to at some point have the script the automation bail out? And I had some peers, actually, that had been
09:50
doing it for a while. They've been UNIX operators, and I could have them do certain things at 5 a.m. in the morning when they came in. I didn't like to come in at 5 a.m. They did, but my automation script would stop at a certain point and have Ted the guy's name was Ted. Do this at 5 a.m.
10:03
When he comes in each morning kind of thing. Implementation, complexity, there's underlying pieces that are very challenging. There's a long decision tree. When we go through provisioning all these different choices, we have to make much less the day to day management, overhead and complexity. How much does that impose?
10:18
And then the automation layer almost being kind of this smear across that has to make up for all these gaps that are often changing. One of my friends now not there anymore. But he used to be the project a product manager around products called like UIO and U. I MP if that means anything to anyone around the operation and provisioning for some of the converged infrastructure products.
10:39
And we would talk about how he was always patching holes that were changing faster than he could fix them with the automation layer. And like every time he took 22 steps forward, it was three steps back, you know, kind of thing. So challenging. This is the environment. 10 years ago, that's changed a good bit, though.
10:55
Yeah, so we wanted to dive into over the next couple of slides, a few things that have changed and a few things pure has brought on and brought to market. That really makes this possible, whereas it wasn't possible before, so api completeness. If you complete this simplicity, you know, we hear all about simplicity,
11:14
the architecture and evergreen, which are really closely related. The how do we achieve balance? What's the What's the foundation of that? And then let's you know, the TCO that you get with pure and we'll see later in the session how that can be enhanced by fusion. Yes, so simplicity is really the underlying storage is a key foundation.
11:37
How do you You can't automate easily without the underlying simplicity. We heard from NTT yesterday in the keynote How simplicity of pure products was a real game changer for them. And a huge effort has gone into and still goes into making everything pure does simple. Another example here is a global bank where they've got over 100 arrays, multiple workloads, and the time they spend on pure storage is the
12:05
equivalent of one FT, and they need 567 times that when they work on other vendors storage. But it's about roughly the same size, the same quantity, more or less. So another foundational aspect that makes, uh that helps with fusion is is evergreen. So evergreen, uh, the architecture is that foundation and the ability to do NDU and
12:29
buy one of the original arrays and and have that upgradable. I love the statistic that 97% of arrays purchased six or more years ago look like new arrays because of Evergreen. The best way that I like to illustrate this Was anyone here at accelerate last year, By any chance? It was a little smaller.
12:47
It was in L A. OK, so Renee from Delta was on stage. And this is the way I've even actually used this with family members of the family reunion, which may speak to how how odd I am when they ask, like, What do you do? And you're trying to explain in some way that's interesting. OK, so Renee was on stage with Prakash being
13:03
interviewed, and he said, You know, I'm gonna go a little bit off script. You can actually see a tweet from Scott Sinclair. He's an analyst. The slightly longer version of that tweet is Renee said, I'm gonna go off script. When I first heard about Evergreen, we could do all this stuff under the covers and the technology and the architecture.
13:16
I thought it was B SI bought in because of the flat and fair maintenance and the guarantee that you wouldn't spike the maintenance in years 45 and six. And then we were able to upgrade 30 day arrays and 30 days with planes in the air. No hiccups. And I believed all that stuff. Now. I didn't necessarily think it was true,
13:35
but I believed it was true. So I sometimes use that to encapsulate. There's a lot of different pieces that go into evergreen. The technology there's, there's pro business and product pieces. There's marketing programmes as well. There's a whole alignment of the financial model that's uniquely enabled by the technology architecture.
13:53
Yeah, so Evergreen is also the foundation for consuming as a service, so you may have heard of Evergreen one. A subscription ever in one subscription allows you to focus on outcomes and service levels that your users users need. Now, Evergreen one is not a requirement for fusion, but together they're really well suited to each other. They're really powerful together.
14:16
There's even a piece here. This was even a conversation with someone earlier this week. This is not the same as a subscription. A subscription is part of it, but this is truly consumption driven as a service. Yeah, another key capability is workload mobility. So I think a lot of people, it's one of the first things people notice about fusion is the
14:36
ability to move workloads. And that leverages the flash array functionality active workload that's built on top of the rock solid technology of active cluster. So the ability and the ability to stretch volumes across arrays, in turn allows mobility between arrays. It's all about the money, and then something else you get with Flash Array and Evergreen is
15:00
a better TC. So this is a simple TC that we see with the typical purchase of Flash Array over six years. And there's two major factors. So Evergreen forever means there's no forklift upgrade and there's no forklift refresh, so you don't have to rebuy your storage like you might have to with anyone else and then up to 85% savings in energy.
15:26
That adds to the overall TCO, depending what the alternative is now. We'll come back to this TC earlier TC later and see how fusion might affect it. So that gives a little bit of a sense, right? Things that were certain ways in the past just by having pure we've actually improved on a lot of the underlying building blocks and capabilities. This isn't diminishing in any way what we're
15:49
literally talking about next with fusion. But we have a better foundation to build on top of that actually makes this more not more possible. It makes it possible at all. Frankly, Section number four, we are moving along intentionally keeping it moving last session of the day. And And I meant to say everyone here gets a
16:07
gold star for coming to the 4 p.m. last session of the day, You can pat yourself on the back. Thank you for being here. You know, you always wonder that when you're like, hm, I'm the last one who's gonna show up. So I appreciate it, but trying to put our your us in your shoes a little bit of if you're sitting at your virtual
16:22
desk, you know, and you're thinking about you know, internal IT external IT questions. And what are the things that relate to that? If I'm kind of managing up within my organisation or externally or even inside, or I'm on the hands on person who's responsible for doing it? So what? We we realised we wanted to kind of try and frame the challenges,
16:41
these two ways of the external, managing up time to market agility. Shadow IT. I've actually at different points in my career. I've been the shadow IT or I've been stepping on the shadow IT right. Sometimes you play different roles, to be honest, if you're if you're technically capable, aligning storage to the business and the financial timing that relates to having a true
17:01
utility model. But do the automation of this fusion allow that, and actually being able to embrace sometimes public cloud concepts about you have providers. You have consumers then looking internally. This is a flip side of some of what we talked about before, thinking I have to deal and think about how do I handle storage silos? How do I handle stranded capacity? As things move around,
17:20
workloads grow and shrink scale is painful, and so is low utilisation. Sometimes it's not our fault, the project said. We wanted to do this thing and we bought all the stuff, but only 10% of it gets used. But look that on your dashboards, right? But how can you handle that in a way that's viable, much less? You even buy something?
17:38
What's the time to value? Does it take six or 12 months to ramp it up and the depreciation schedules ticking along during that time frame right kind of thing. So what we realised is we thought about we could we go through and for each one of these, give the fusion answer for each one of these. And then we realised it would be way too long of a session, and someone earlier said you might already have too many slides,
17:57
so we didn't take that approach because we're pretty sure it wouldn't work. So what we wanted to do for the rest of this section is walk through an overview of fusion and call out core architectural concepts and features that relate to helping with these challenges. And if you're watching along, we're betting we're hoping you'll be able to put it together. We might have a quiz at the end.
18:16
Actually, not because then you would walk out. We're actually going to walk through some of the features and capabilities here. I do want to make sure to highlight that. At the end, we will have time for questions and answers and really appreciate actually the trust, uh, in place in and myself by Larry, who's sitting in the back there.
18:32
I don't know if he's trying to be anonymous or not, but he is the product manager for for Fusion, for pure fusion. Um, as even as we care about these topics intensely from personal background, Um, he lives and breathes this every single day, so he's here as well. So we may. We might just pull him into the Q and a
18:46
depending on how hard the questions are. So what is fusion? There are a couple of different ways that I'll define it. Even aside from what's out on this slide, sometimes I'll say it's an out of band management layer. It's not in the data path,
19:01
but, like I said there, self service, autonomous storage as code built for limitless scale that might have just checked all your buzzword bingo cards. But buzz words used appropriately are good. All right, so let's unpack that maybe starting even on maybe even starting on the top. Right? Instant self service provisioning providing the
19:21
ability for end users consumers to provision instantly, not in days or weeks or months. Scale out that goes. Maybe it's near infinite, but goes far past what any individual storage frame could ever do from anybody pure or any other company. Right, enabling that we walk into intelligent workload management. Workloads change.
19:42
I think someone laughed at the joke earlier about requirements Always change, or they always stay the same, right? So it's the initial requirements. They mutate over time. We don't know what that's going to be. We need to have enough flexibility to be able to handle that and then aligning to a cloud operating model. This even goes to some of the idea of bringing
19:58
cloud concepts on premises. Don't leave anything out there or should I keep going? Keep going. Oh, well. Providers and consumers will cover those aspects. I will go So first concept here. So that's the overview of fusion.
20:12
Kind of what it does. Providers and consumers is the idea that often the people administering managing the storage provisioning essentially have to be the same as the people, consuming it, practically speaking right kind of thing. So the idea here is almost to if you take let's say that you're in the public cloud. You understand this very intuitively. You consume a service,
20:33
you are a consumer. I'm going a little bit out of order. You know, you maybe an application owner and an operator. You want to consume services, you care about applications and business value, and you don't want to know or care about what is provided on an underlying basis as long as the policies and the S and the things you need are provided in that service interface.
20:51
But there has to be a provider who is thinking about those things classically an IT, or storage engine engineer manager. Inside the IT organisation, they're the ones defining the services in pure fusion. They don't necessarily own the applications, although hopefully there's good, you know, internal politics and you know each other and get along.
21:07
You're aware of what you're building for, but you care about the platform characteristics, right? The network and the policies that we'll get into in just a minute so two major constituencies, and that's actually baked into not just a slide. It's actually baked into how how fusion operates from a day to day basis.
21:26
OK, so we'll dive into a little bit about some of the fusion concepts here. So what's an availability zone? And what's a tenant in fusion? So think of an availability zone in fusion as the storage system, so you're no longer looking at individual arrays under the covers. That storage system or availability zone is built out of multiple arrays that are all
21:50
reachable by the same hosts and all connected to the same back end network, so you can consider them as one array. Um, how key points as to how fusion is different? Well, the providers, the storage team now known as provider, they no longer manage individual arrays. And those providers,
22:10
they no longer provision volumes for their consumers or the users they no longer attach them to hosts. Consumers make their own volumes by working with what we're calling tenant and tenant spaces. And this is the R back concept within fusion rob based access control. Thank you. Uh, and with an API key as a tenant,
22:34
you can then provision, uh, volumes and manage the storage just within your tenant space. Um, so we have storage classes, protection policies, um, and storage policies that Andrew is gonna cover. Now, I think even there's a little bit more here from a visualisation standpoint. So before Andrew does that, um, yeah. So if we look at a data centre and this is
23:00
really a topology attempted a topology, So a number of data centres you've got individual arrays and racks rather than do that with fusion, um, you would group those into availability zones and regions, and then those are the concepts that that fusion would deal with or your users would deal with. Uh, I'm curious.
23:23
For how many folks is this availability zones? A brand new concept, maybe show hands, Maybe a little. OK, this is very similar to public cloud concepts, right kind of thing. It is an inherent that couldn't do it with only with public cloud right kind of thing. That's why we're talking about it on Prem premises. Um, but this is in some ways what you would see
23:41
with a public cloud provider, especially around availability zones, tenants that might be a little bit different, but we know that you may need to enable multiple tenants multiple groups inside your organisation Next. When we say policies, it has to mean something. So what does it mean for fusion? There are core things that we do want to expose as almost the selection choices that we make at
24:03
an opinion level, as the as the storage team or the providers. And then there are ones that we may want to provide options to the consumers around. So there are likely going to be multiple policies. You could do it with one policy if you want, but multiple policies around storage services and classes so that those go into things like latency or availability.
24:22
It even goes into things like I ops and bandwidth limits sometimes some of the classic things that you might think about in other times when you're provisioning storage. OK, but that needs to be. There may be multiple options because there may be multiple underlying tiers inside the storage landscape that have different latency or different capacity capabilities.
24:39
Let me get into hosts, so this is pretty straightforward as well. Uh, we need to control who accesses stuff, especially in a tenant environment. So which hosts can access? What, what piece of the environment and the protection policies? This gets a little bit more if you're familiar with this. Is built in some ways on snapshots from an
24:55
underlying peer standpoint. Better than average snapshots, not the challenging snapshots of yesteryear, but express to the consumer in terms of what they are expecting about on SLA. So this starts to get into the frequency of protection the retention of the protection period right and expressing it in ways that a consumer could understand. They don't necessarily need to understand how
25:14
it happens under the covers. For them, it's magic. Maybe there's a little bit of magic in how really good snapshot made of data is handled right, but they can see it that way. Let me talk about placement groups so within a tenant, so a tenant might be an internal customer. Multiple business units, multiple applications.
25:30
Within there you may have the need for various control around the placement. Who here is familiar with VMware? DRS Storage, V motion, et cetera. I'm getting some nods, or at least mostly, I want to assume everybody, but these are sometimes familiar concepts. Basically, if you actually understand the application a three tier application individual,
25:48
some things really need to live together. That would be affinity. They need to be high speed connected. Some things really need to stay apart for resiliency. It doesn't help to have a scale out Web tier if all the Web servers end up living on the exact same spot or the storage right kind of thing, even consistency of snapshots packaging up that it's an application being represented with
26:05
multiple pieces and components that all need consistency and then even rebalance granularity will continue deeper into rebalance but that as things are rebalanced, it needs to follow a policy where they stay roughly together and within the same pieces. So there's some thoughtfulness here about what might need to happen at a policy level and be expressed so that we can make it automated underneath Ah,
26:30
and then play this out. So let's let's actually just do a little bit of a scenario. So a little bit of a heavy slide. I realise I'll narrate you through it a little bit. And we are. I think we're roughly still on time, way into the Q and a just a little bit.
26:42
But so, uh, so we have tenants at the top consumer. You see it in a provider. So say from a consumer standpoint, you might want production and reporting copy in the same availability zone. You want that to live together, that could be accomplished. Or maybe you want the production and copy Separated.
26:57
You want them in adjacent availability in adjacent tenant spaces, but within the same within a different availability zone. And actually you can see So the orange maps down there to the orange, the black maps to the black. I think we pulled out a couple of the animations. Or maybe the tenant and the production and the
27:12
copy need to live in totally different regions so you can actually see under the covers how that's actually split out between the green and Region one, and then it's over. In Region two, we can do load balancing within the same a Z for performance. Or we might do load balancing for availability within within and do that between availability zones.
27:32
So these are the kind of scenarios and the colours at the top map the colours at the bottom, sometimes I like to take twice as long in this slide, but we'll keep moving because you can ask questions. If not, that does not want to go. There we go. Last piece back to you.
27:46
OK, so this is showing a rebalance scenario. So where you, um, where you change requirements? Or maybe one array is more full than another, or you add an array to the environment. The the Fusion will take care of the back end, rebalance and keep that invisible to the users at the front end. As we click, we'll just click through the magic
28:13
of transitions. It's expanding, it's growing between arrays. And then I think, magically, Hey, it lives over there. This is where you know that it's true because you saw it on a slide. But it is actually true under the covers as well. So wrapping up the fourth section, hopefully, this gives a sense.
28:34
We realised that we would only go so far in the section. We wanted to set the landscape, as we think about the two types of challenges silos, stranding, capacity, scale that maps into a bunch of the policies and pieces in the and even rebalance and workload, making it faster to request and provision actually aligning the consumption up into the hands of the consumer. Frankly, a lot of the reasons that people go to
28:55
public cloud, at least from a storage standpoint bringing that on Prem for the final section, though there has to be a financial aspect back to you. Thank you. Yeah. So there are a number of areas where fusion helps and we could look at quantifying value and certainly with all the customers I deal with, they want to They want the numbers. They want to quantify the value,
29:16
understand the TCO, help make the internal business case before they deploy something. So we thought about this, and one thing you could look at is simplification of speed of development. How quickly is it to deploy an internal self service? If you're programming to array, APIs versus the fusion api. So that's one thing you could look at.
29:39
The other is what are the savings based on safely maintaining a better utilisation of assets? The third thing I thought you could look at is you could explore the agility of being able to offer a new tier of service, So let's say you wanted to offer a Tier two service and deployed a Flasher AC into your environment. What's the benefit of your users?
30:02
Being able to select that new service and have fusion retire and move the workloads automatically, or what we're going to dive into a bit more is how can the automatic rebalance capability solve some of those utilisation challenges that you typically get in an environment? So we're going to dive a little bit deeper into that and what we see.
30:24
I'll click. So we looked at a large, pure customer, and this large, pure customer is offering an internal cloud service to their users. And when we look in pure one, we see you know, you can look in the pure one capacity plan of view. What is the growth that we've seen over the last 365 days? And it's really slow, so we see some examples.
30:49
We see it takes more than a year to reach 50% utilisation. We see some arrays that are over two years old that have less than 50% utilisation and also uneven utilisation, so you might have two arrays in an environment and one of them is 75% full, or 75% busy, versus 5% for a newly deployed array. And this problem is driven by the unpredictability of growth when deploying a
31:18
service. So despite really diligent capacity forecast by this customer, the that growth and demand didn't materialise as predicted. And this pattern tends to repeat itself in many large organisations that have fleets of arrays. And this problem is actually getting worse, with some of the agility being offered at the application level or that application layer. So think of port Works.
31:44
For example, Port Works gives you ability to move workloads between infra infrastructure stacks. Be that on Prem or between on Prem Infrastructure and the Cloud, so you don't actually know over time where workloads and where capacity is going to end up. So that unpredictability makes predicting infrastructure requirements even more unpredictable.
32:07
Now, sometimes when it's if you're lucky that low utilisation is down to getting fantastic data reduction on the flash array, I guess that's a nice problem to have. But still, how can we solve for that low utilisation issue? So one way is an as a service model, So we talked a bit about evergreen wine earlier, and that would allow you to pay for just effective used capacity.
32:33
And in that model, pure ensures. There's always sufficient head room for you to grow into, but that's only really half the story. That's only half the solution because you're still paying for. You're still having pain in terms of energy use in terms of capacity that's deployed that maybe shouldn't be there. So let's look at a scenario.
32:55
If we look at a typical deployment scenario without fusion, you might install arrays. So in this particular customer situation, they'll deploy a pod of arrays for a service, and their capacity planning showed they needed three arrays. So they deploy for that end state because it's just too difficult to deploy less than three. They want to provision volumes where they're
33:19
going to end up, and the performance ramp up is really slow, and you can obviously expand physical capacity non disruptively with flash array. But then utilisation can remain low for a long time, as we saw now, an alternative approach is if you then if you created if you deploy fusion defined your tenant spaces defined your availability zone here we've provisioned the same volumes to the
33:47
users have provisioned the same volumes that we saw earlier. But physically, they're just they're on a single flash array in that availability zone. And so you've installed the first array, uses provision capacity, and then, when needed, when demand dictates, when utilisation grows, add the second array and fusion Rebalances the workloads.
34:14
And then again we could add a third array, which would mean a further rebalance. Now, what would that do to the TC? So we saw this earlier. A typical evergreen TC for a capital purchase. Um, so with with that model we just saw it would mean shifting the cost to the right. So I guess a 60% in year saving for year one who doesn't want that and really
34:42
making sure that you can. You can deal with that unpredictability and that unpredictability of demand that we talked about. And you've obviously got the time, value, benefit of time, value of money, benefit as well. With that, we are just about done.
35:02
Thank you for staying with us, I think. Amazingly but I, I think my one goal was to keep everyone awake to happy hour and I. I think we just about accomplished that, but to bring it down to wrap up. Hopefully, this makes sense kind of why we want to go through this slow, not just start with OK, here. Here's fusion and here's why.
35:17
We think it's wonderful. But actually how this relates more to you as well. And if you remember the previous picture there was, there was a nice little missing spot in the orange box there. So how you say this? This is where pure fusion fits as that management layer that can provide these benefits, of course,
35:34
with pure arrays. Want to make sure this is a out of band management construct? It is not living in the data path, and it is specifically for pure products, right, as you can see as you can see there providing capabilities all the way up the stack. And I even love what you were walking through that alignment of value to what the business cares about.
35:52
The business would say, You know, I'm not ready. I'm only doing this stuff in year two and three, but I got to spend all the money in year one. We can help that with a financial construct. Evergreen one and alignment there. Consumption base. We can help with technology pieces. This lets us do outcomes. I'm not going to read this at you.
36:07
We left this in here so you have it as a follow up. But if you're speaking to a strictly business focused audience, the outcomes that you could talk about self service speed much faster time to provisioning dramatically higher utilisation with lower costs. Aligning that to the business initiatives and the value of the business sees consistent
36:25
operation. We, as humans are often the weakest link right. So automation avoids inconsistent operations, especially if it's built on a foundation that's not too fragile to be able to handle that. And then, of course, be able to scale dramatically, faster when needed, or or scale faster, but also scale when needed.
36:43
If you'd like to learn more, please feel free to visit pure storage dot com slash fusion. And if you would like to test drive the fusion API experience, you can actually do that right now. There's a QR code. We'll leave it up there. Larry, I don't know if you have any of the of the cards that you were giving out earlier.
36:59
With those, you might have them. So awesome. So, um, see you later to get your card with the QR code there. But if you'd like to actually try this out because this is meant to be an API first experience overall. Yeah, this just went live yesterday. I'm keen to go try it out, but, uh, so it's brand new.
37:13
Uh, take a look at it. And with that, I think we are finished with the official part of the presentation for the audience of the future. Who is listening to us? Thank you so much for listening for the audience of today. Thank you for being here For those who are here live.
37:29
You get to ask questions. Now, folks, in the future, they have to do it offline.
